^{page 56 note (1)} G. A. Renton, ‘Forecasting British exports of manu factures to industrial countries’, National Institute Economic Review no. 42, November 1967, pages 35-51; A. H. C. Broadbent, E. A. Anyanwu, and N. C. Garganas, ‘Forecasting trade between OECD countries and primary producers’, National Institute Economic Review no. 52, May 1970, pages 46-51. M. J. C. Surrey, ‘The analysis and forecasting of the British economy’, NIESR Occasional Papers XXV, CUP, 1971, includes a briefer exposition and also describes a method developed in the National Institute of forecasting British exports in the aggregate by the use as explanatory variables of the index of industrial countries' manufacturing production and the ratio of the index to its trend. Primary producers' reserves and export prices, pressure of demand in the United Kingdom, and prices of British exports compared with those of competitors were found not to be statistically significant in this aggregate relationship.

^{page 57 note (1)} The question may arise whether the equations estimated here could provide a useful basis for calculating export pre dictions after United Kingdom entry into the EEC. While the regression results described here suggest that changes in EEC tariff discrimination against British exports, as measured by the relevant proxy variable, have not been a significant influence on the various categories of exports, it may still be desirable in preparing the forecasts for the period after British membership of the EEC to allow for the possibility of some effects of the tariff and cost changes. For this purpose the estimated price elasticities may provide a useful base.

^{page 57 note (2)} It can be shown, in fact, at the micro-level at least, that under certain conditions variations in domestic demand pres sure can be associated with either an adverse or a favourable change or even with no change in exports, the actual outcome depending upon the underlying behaviour of businessmen, the structure of the home and overseas markets, and the cost conditions under which firms operate.

^{page 58 note (1)} For example, R. J. Ball, J. R. Eaton and M. D. Steuer (‘The relationship between United Kingdom export perform ance in manufactures and the internal pressure of demand’, The Economic Journal, September 1966) found that fluctuations in Britain's share of exports around its downward trend during 1954-64 were related inversely to cyclical movements in domestic capacity utilisation as measured by the ratio of United Kingdom industrial production to trend. Their work on the machine-tool industry (‘The effect of waiting times on foreign orders for machine tools’, Economica, November 1966) presents evidence which is consistent with their inference for manufacturing as a whole. A similar conclusion was reached recently by J. R. Artus (‘The short-term effects of domestic demand pressure on British export performance’, IMF Staff Papers, 1970) for British exports of motor vehicles and chemicals during the period 1960-67. On the other hand, J. D. Smyth (‘Stop-go and United Kingdom exports of manufactures’, Bulletin of the Oxford University Institute of Economics and Statistics, vol. 30, 1968), using annual data, found that it was the rate of change of unemployment, rather than its absolute level, that had influenced British exports of manufactures during 1954-64. R. A. Cooper, K. Hartley and C. R. M. Harvey (‘Export performance and the pressure of demand: a study of firms’, University of York Studies of Economics, George Allen & Unwin Ltd, London, 1970) have recently looked at the performance of firms in a limited number of British industries—pottery, motor cycles and cycles, office machinery, and domestic electrical appliances— in relation to variations in internal demand pressure over the period 1958-66 on the basis of time series and interview data. They report that, on the whole, firms thought their exports were unaffected by changes in home demand. Their regression results have been variable but in a number of cases they provided little or no support for the pressure hypothesis. N. C. Garganas (‘A study of the effect of internal demand pressure on the United Kingdom balance of trade in manu factures, 1950-1966’, unpublished doctoral dissertation, University of London, February 1971) investigated the supply effect of fluctuations in domestic demand pressure on both the export and the import side of the United Kingdom balance of trade, using quarterly data over 1950-66 and for a shorter time period. After testing with a variety of specifications, he found that neither domestic nor foreign pressure on capacity as measured by alternative indicators had been a significant influence on exports. These findings were confirmed by the evidence obtained for individual industries.

^{page 58 note (2)} A fuller description of the variables and their sources is given in the Appendix.

^{page 58 note (3)} These groups correspond as follows with the SITC (numbers in brackets representing their respective shares in total exports to EEC in 1970): non-manufactures (17.0), sections 0-4 and 9; chemicals (9.2), section 5; metals (12.3), divisions 67-9; other semi-manufactures (12.9), rest of section 6; machinery (28.2), divisions 71-2; transport equipment (11.7), division 73; miscellaneous manufactures (8.7), section 8. It was expected that exclusion of ships and aircraft (groups 734-5) would give better results for transport equipment. But although a more plausible price elasticity was obtained in this way (see the footnote on page 62), the prediction error for 1970-71 was in fact increased.

^{page 58 note (4)} The only existing separate index of United Kingdom export unit values for a particular market is the series derived by the Board of Trade (now Department of Trade and Industry) for total commodity exports to the sterling area countries over the period 1949-60. There were wide discre pancies between this and the corresponding unit values for all exports from 1949 to 1953. In the subsequent years, however, the two movements showed a tendency to coincide except in 1957.

^{page 59 note (1)} The price indicator for competitor countries was con structed in two stages: first a price index was calculated for five major exporters to each EEC country separately weighted by each supplier's share in that market; then a weighted average of the resulting price indices was obtained for each commodity group, the weights being related to the relative importance of each EEC country in United Kingdom exports.

^{page 60 note (1)} Since the volume of exports and their prices (and possibly also overseas expenditure or output variables) are to some extent jointly determined endogenous quantities, the use of single-equation methods may, of course, be expected to lead to some estimation bias. Some trial attempts were in fact made to work with instrumental variables. It was not possible within the limits of our resources to pursue this aspect in depth. It is, however, worth noting that the trials carried out suggested much higher price elasticities. The topic deserves fuller treatment but the difficulties of correctly specifying other relevant relations in the system (or selecting the correct instrumental variables) will be appreciated.

^{page 60 note (2)} In the case of metals the observation for 1957 I seemed abnormal and was excluded.

^{page 60 note (3)} About 80 examples of the other equations are available on application to the Editor, National Institute Economic Review, 2, Dean Trench Street, London, SW1.

^{page 60 note (4)} M. C. Fessey, ‘Short-term forecasting of United King dom exports’, Economic Trends, May 1967, Appendix 2.

^{page 62 note (1)} See for example, R. L. Major and S. Hays, ‘Another look at the Common Market’, National Institute Economic Review no. 54, November 1970. (Other literature on this subject is listed in Section II of the bibliography published in National Institute Economic Review no. 57, August 1971, pages 58-9.)

^{page 62 note (2)} And when ships and aircraft were excluded (see the footnote on page 58) the price elasticities were reduced to about —1.9 and −1.8 respectively. On this basis the best equation, after adjustment for serial correlation as in table 1, was: lnX = −6.534 + 2.029 InCE - 1.790 ln P_uk/P_w–₁ + (0.294) (0.801) ^0.386D_(0.094)4 ρ = 0.741 R² = 0.64 SE = 0.11 DW = 1.87

^{page 63 note (1)} The objections to basing forecasts on models with serially correlated errors are twofold. First, they would be biased, since no account would have been taken of the recent values of the errors. Secondly, they would have large sampling variances, having been derived from inefficient least squares estimators.

^{page 63 note (2)} Durbin's method, which is less laborious and no less efficient than most of the others now available, can be sum marised as follows: consider the simple case Y_t = bX_t + u_t in which errors constitute a first order autoregressive process u_t + ρu_t-1 = et where e is assumed to be independently distributed and to have a constant variance. The first step is to estimate a linear regression of Y_t on Y_t-1, X_t, and X_t-1. Let p denote the estimated coefficient on Y_t-1. The second step is to calculate the series Y′_t = Yt — p Y_t-1 and X′_t = Xt - ρX_t-1; then regressing Y′_t on X′_t yields an estimate of b. See J. Durbin, ‘Estimation of parameters in time-series regression models’, Journal of the Royal Statistical Society, Series B, vol. 22, no. 1, 1960.

^{page 64 note (1)} See Shirley Almon, ‘The distributed lag between capital appropriations and expenditures’, Econometrica, January 1965.

^{page 65 note (1)} The predictions from equations corrected for serial correlation of residuals were made in two steps; first, an estimate of exports was obtained for each period from the transformed equation using observed values of the exogenous variables; then the predicted value was adjusted for the estimated error in the preceding quarter using the equation: u_t = ρ u_t-l + e_t.

^{page 68 note (1)} National Institute Economic Review no. 59, February 1972, page 11, table 4.